Tumor necrosis factor α (TNFα) is a cytokine that plays a critical role in the regulation of inflammatory processes, both idiopathic and infectious. In the case of infection, one of its key roles is to facilitate cell-to-cell communication in the control of invasive infection. As might be expected, there is increasing evidence that inhibition of TNFα is associated with the development of serious infectious diseases (1) and difficulty clearing infections once they develop (2). At the same time, this treatment strategy has resulted in disease control in certain inflammatory diseases including rheumatoid arthritis (RA), Crohn's disease, psoriatic arthritis, and juvenile rheumatoid arthritis. The spectrum of pathogens causing invasive disease in patients receiving TNFα blockade therapy ranges from common gram-positive and gram-negative bacteria to more opportunistic organisms such as Mycobacterium tuberculosis, Cryptococcus, and Aspergillus (Table 1). The extent of the infections ranges from localized to disseminated. It appears that not only is the incidence of certain infections increased with anti-TNFα therapy, but the ability to contain these infections is also impaired. This underscores the need to develop measures to reduce the risk of infectious complications associated with TNF blockade.
Table 1. Infectious agents associated with tumor necrosis factor α inhibition
| Mycobacterium avium-intracellulare (1)|
| Mycobacterium tuberculosis (13)|
| Streptococccus pneumoniae (2)|
| Listeria monocytogenes (34)|
| Candida albicans (1)|
| Pneumocystis jiroveci (1)|
| Aspergillus fumigatus (14)|
| Histoplasma capsulatum (41)|
| Cryptococcus neoformans (44)|
| Coccidioides immitis (52)|
The purpose of this review is to delineate the role of TNFα in host defense against infection, define the effects of TNFα blockade in clinical sepsis trials and in animal models of infection, and, most importantly, describe the variety of infections that have been reported in patients receiving anti-TNFα therapy. Recommendations aimed at preventing infectious complications and treating established infections associated with anti-TNFα therapy will be proposed. In the absence of any established controlled clinical trials designed to investigate strategies to limit anti-TNFα–associated infections, it is necessary to propose guidelines based on case reports, case series, anecdotal experience, and expert opinion. For other groups of immunocompromised patients, the concept of the therapeutic prescription (3) has been advocated, and it has been shown to be effective in the arena of transplantation. The therapeutic prescription has 2 components: an immunomodulating effect to control the underlying condition and an antimicrobial program to make disease-modifying therapy safer. Formulation of such a program requires a clear understanding of the host defense defects that are produced and the observed clinical infections that have occurred.
TNFα is a cytokine within the TNF superfamily which acts as a central mediator of inflammation and immune regulation. It is synthesized and secreted chiefly by macrophages in response to proinflammatory stimuli such as bacterial lipopolysaccharide. It is expressed as a transmembrane protein and is functional on the cell surface, or, if cleaved by a specific metalloproteinase, is released from the cell surface as a soluble homotrimer. TNFα binds to cell surface receptors, TNF receptor I (TNFRI), and TNFRII (4).
TNFα antagonists have been demonstrated, in placebo-controlled clinical trials, to be highly effective in the treatment of certain inflammatory diseases such as RA (5–9), psoriatic arthritis (10), juvenile rheumatoid arthritis (11), and Crohn's disease (12)—disorders in which TNFα may have a significant role in pathogenesis. As of January 2003, 3 TNFα antagonists, the monoclonal antibodies infliximab and adalimimab and the p75 TNF soluble receptor etanercept, have been approved by the Food and Drug Administration (FDA) for clinical use in the US.
In clinical trials of these agents, an increased rate of serious infections was generally not noted. It was not until infliximab and etanercept were approved for and used in clinical practice that reports of sepsis, as well as tuberculosis (TB) and infections with atypical mycobacteria and other organisms, were observed. In 2001 an FDA advisory committee reviewed toxicity issues related to TNF inhibitors, specifically, infections that included TB, endemic fungi such as Histoplasmacapsulatum and Coccidioidesimmitis, yeasts such as Cryptococcus neoformans and Candida species, Pneumocystis pneumonia, molds such as Aspergillus, and bacteria such as Listeria monocytogenes (1).
Role in host defense
The biologic effects of TNFα on immune cells are remarkably broad (Table 2). Since TNFα plays a major role in monocyte, neutrophil, B cell, and T cell function, as well as in facilitating necessary communication among these cells, it is not surprising that the neutralization of TNFα has led to infections associated with impairments in the function of these cells. The increased risk of TB (13) and, in particular, extrapulmonary TB is reminiscent of findings in patients with depressed cell-mediated immunity (e.g., acquired immunodeficiency syndrome). Invasive aspergillosis (14) associated with anti-TNFα therapy has been described and may correlate with neutrophil- and/or cell-mediated immune dysfunction. Uncontrolled pneumococcal necrotizing fasciitis (2) as a result of anti-TNFα therapy may relate to phagocytic dysfunction.
Table 2. Biologic effects of tumor necrosis factor α (TNFα) on immune cells*
| Autoinduce TNFα|
| Chemotaxis and migration|
| Inhibit differentiation|
| Suppress proliferation|
| Increase phagocytosis|
| Increase production of superoxide|
| Stimulate integrin response|
| Induce T cell colony formation|
| Induce superoxide in B cells|
| Induce apoptosis in mature T cells|
| Activate cytotoxic T cell invasiveness|
TNFα inhibition and sepsis
TNFα, in conjunction with other proinflammatory cytokines (i.e., interleukin-1), is an important mediator of sepsis. After the introduction of bacteria or endotoxin into the systemic circulation in animal models, the concentration of TNFα increases, and the effects of administration of TNFα in humans reproduce the physiologic and chemical changes seen in severe sepsis (15). Furthermore, antibodies directed against TNFα have exhibited a protective role in animal models of sepsis (16–18).
These observations led to clinical trials of anti-TNFα therapy in sepsis. In a study of the efficacy of etanercept in 141 patients with septic shock, there were 10 deaths among 33 patients in the placebo group (30% mortality), 9 deaths among 30 patients receiving low-dose etanercept (0.15 mg/kg body weight) (30% mortality), 14 deaths among 29 patients receiving intermediate-dose etanercept (0.45 mg/kg body weight) (48% mortality), and 26 deaths among 49 patients receiving high-dose etanercept (1.5 mg/kg body weight) (53% mortality) (19). There were more gram-positive infectious causes of sepsis in the 3 etanercept groups compared with the placebo group, in addition to an increase in Pseudomonas aeruginosa infection in the group receiving 0.45 mg/kg of etanercept (P = 0.08). In 2 other trials using TNF monoclonal antibodies for the treatment of sepsis, there was no significant difference in all-cause mortality at 28 days in patients who received placebo as compared with those who received TNFα monoclonal antibody (20, 21). These clinical data represent prima facie evidence that despite the role of TNFα in the augmentation and pathogenesis of sepsis, neutralization of this cytokine does not lead to improved survival; in fact, in the case of higher-dose etanercept, an increase in mortality was noted.
Infections associated with TNFα blockade
TNFα plays an important role in the host defense against M tuberculosis. Kaneko et al (22) created a TNFα-knockout mouse model and demonstrated that challenge with Mtuberculosis decreased the survival time of these mice from 50 days to 33 days. At necropsy, the mice had diffuse abscesses in the lungs, liver, spleen, and kidneys. Histopathologically, the TNFα-knockout mice exhibited numerous necrotic regions filled with acid-fast bacilli (AFB) in multiple organs. No typical granulomas were identified in the necrotic areas. The amount of AFB within the various organs was significantly higher in the TNFα-knockout mice compared with wild-type mice. TNF messenger RNA can be detected in granulomas in bacillus Calmette-Guérin (BCG)–infected mice (23). Furthermore, injection of anti-TNF IgG antibody significantly reduces the development of granulomas, which are mycobactericidal and play a major role in bacterial containment and elimination. TNFα is involved in host defense against atypical mycobacteria as well. Mice that were genetically deficient in TNFR p55 developed necrotic granulomas when infected with Mycobacterium avium (24). The infection was uniformly fatal in these mice, whereas infected mice with normal TNFR p55 survived for the duration of the study.
In a randomized phase III trial comparing infliximab with placebo treatment in RA patients who were receiving methotrexate, 1 case of TB was reported (5). Seventeen cases of infliximab-associated TB were originally reported from Europe (11 reports) and the US (6 reports), prompting a new TB warning in the infliximab package insert and a recommendation that patients be screened and treated for latent TB prior to treatment with infliximab (1). After drug approval and widespread clinical use of infliximab, 70 cases of TB following infliximab treatment were reported to the FDA as of June 2001 (13) (Table 3). Most of these cases occurred within 12 weeks of the initiation of infliximab therapy. Twelve of these patients died, and at least 4 of the deaths could be attributed to the infection. The authors projected that the background rate of TB in patients with RA was 6.2 cases per 100,000, while the rate of TB among infliximab-treated RA patients in the US was 4-fold greater (24.4 cases per 100,000). Over the next 6 months, 47 additional cases of infliximab-associated TB were reported, increasing the total to 117 cases (25). At that time point, >147,000 patients had been treated with infliximab.
Table 3. Tuberculosis associated with infliximab, a tumor necrosis factor α (TNFα)–neutralizing agent*
|Deaths||12 (17) (4 directly related)|
|Ages, median (range) years||57 (18–83)|
|Weeks of TNF blockade treatment prior to tuberculosis diagnosis, range||1–52|
|Cases of tuberculosis with ≤3 infusions||48 (69)|
|Cases of isolated pulmonary tuberculosis||30 (43)|
|Cases of extrapulmonary tuberculosis||40 (57)†|
|Receiving immunosuppressive agents|| |
|History of tuberculosis infection or disease||8 (11)|
|From countries with high incidence of tuberculosis||6 (8.6)|
|From countries with low incidence of tuberculosis||64 (91.4)|
Several important observations have been made regarding TB in the setting of anti-TNFα therapy, including the following: 1) the incidence of active TB was increased with the use of infliximab; 2) the majority of patients had extrapulmonary TB (57%), in contrast to TB in immunocompetent patients, of whom 18% develop extrapulmonary disease; 3) almost 25% of these patients had disseminated disease, whereas disseminated TB typically accounts for <2% of cases of TB in human immunodeficiency virus (HIV)–negative individuals (26); 4) there were a large number of biopsies used to establish the diagnosis of TB, which may reflect an increase in atypical manifestations of TB in patients receiving TNFα blockade treatment. Granulomas were not observed on histologic specimens in the setting of active disease, consistent with the findings in the knockout mouse models.
In contrast to findings in patients treated with infliximab, in US and European trials of the p75 soluble receptor etanercept (2,024 patients), no cases of TB have been observed (1). However, by 2002, 25 cases of TB in patients receiving etanercept had been reported through the MedWatch reporting system (27). In 13 of these cases (52%) the patients had extrapulmonary disease, and there was 1 death directly related to infection.
Thirteen cases of TB among 2,468 patients receiving adalimumab in the clinical development program for this drug have been reported (28). Seven of these cases of TB occurred early in the clinical trials. Since the majority of those 7 patients had findings consistent with TB on baseline chest radiographs, the FDA recommended TB screening and treatment of TB if present, for all patients prior to study enrollment.
One clinically important distinction between the TNFα inhibitors is that the median time to onset of TB differs markedly. In patients treated with infliximab, the median time is 12 weeks, whereas in those treated with etanercept, it is 46 weeks. The median time to onset of TB in the adalimumab clinical program is 30 weeks. Despite the differences in the number of cases of TB observed with these 3 agents, it should be noted that disseminated TB has been seen with all 3. Currently, there is no clear immunologic explanation for the differences in the incidence of M tuberculosis infection between the various treatments, since they all neutralize TNFα. Pharmacologically, there are important differences. In patients receiving infliximab, the level of TNFα neutralization is sustained for weeks due to slow clearance of IgG from the circulation. The duration of TNFα neutralization is significantly shorter with etanercept. It is possible that the sustained neutralization of TNFα with infliximab or adalimumab may place the patient at greater risk for opportunistic infections such as TB, compared with the risk associated with the more intermittent peaks and troughs of neutralization occurring with etanercept (29).
In murine models of pneumococcal pneumonia, mice infected intranasally with S pneumoniae had increased TNFα levels in the lungs and serum, which were proportional to the increased bacterial burden in the lungs (30). The administration of anti-TNFα increased both the microbial burden and mortality. Five of 15 control mice died after infection with S pneumoniae, whereas 11 of 15 anti-TNFα–treated mice died after infection. These results suggest that TNFα helps prevent bacteremia and death. There were significantly fewer circulating neutrophils in the blood on day 7 after infection in the anti-TNFα–treated mice compared with the control group. In another study, anti-TNFα–treated mice had 4-fold more S pneumoniae colony-forming units recovered from the lungs than control mice 40 hours after intranasal inoculation (31). Furthermore, anti-TNFα–treated mice died of pneumococcal pneumonia significantly sooner than control mice. Moreover, in a murine model of pneumococcal peritonitis, 100 colony-forming units of S pneumoniae produced fatal peritonitis in TNFα-deficient mice, but not in wild-type mice (32). TNF deficiency also led to a substantial increase in the level of pneumococci in blood and the spleen.
A case of fatal sepsis and pneumococcal necrotizing fasciitis in a patient receiving etanercept has been reported (2). Another case of fatal pneumococcal sepsis and necrotizing fasciitis in a patient receiving long-term etanercept therapy has occurred at our institution. Pneumococcal necrotizing fasciitis is exceedingly rare in immunocompetent hosts, and its presence suggests a compromised host defense, which in these cases is likely due to TNFα inhibition.
Systemic TNFα release occurs in mice challenged with a lethal dose of L monocytogenes. In mice infected with a sublethal dose of L monocytogenes and then treated with anti-TNF IgG, the sublethal infection is converted to a lethal one (33). Histologic examination of livers and spleens infected with L monocytogenes reveals that control mice exhibit granulomatous inflammation with small numbers of bacteria present, whereas mice treated with anti-TNF IgG have a large bacterial burden, few mononuclear cells, and little granulomatous inflammation. It appears that TNFα plays a role in the formation of granulomas, an important mechanism in limiting growth of intracellular pathogens.
Under normal conditions, immunologic memory follows nonlethal Listeria infection. All control mice administered an immunizing inoculum of Listeria 27 days prior to receiving a lethal dose of Listeria survived the infection. In contrast, mice immunized with Listeria followed by anti-TNF IgG administration died 4 days after challenge. Importantly, administration of exogenous recombinant TNF to mice prior to lethal Listeria challenge prolonged survival times. The livers and spleens of mice administered recombinant TNF prior to Listeria infection had significantly less infection after 24 hours.
As of December 2001, 15 cases of listeriosis in the setting of TNF blockade had been reported to the FDA (34). Fourteen cases were associated with infliximab therapy and 1 with etanercept therapy. The majority of these patients were age 60 years or older and were also receiving at least 1 other immunosuppressive agent. Fever, fatigue, headache, and confusion were the most common symptoms, occurring within 4–290 days of receipt of the first dose of infliximab. Five patients treated with infliximab and 1 treated with etanercept died of Listeria sepsis. In an addendum to that report, the authors noted that since submission of the manuscript, an additional 11 cases of listeriosis had been reported to the FDA, of which 10 were associated with infliximab therapy and 1 with etanercept therapy. Two of the infliximab-treated patients had died of Listeria infection. Two cases of systemic listeriosis associated with infliximab therapy, both occurring in the setting of cholecystitis, leading to meningoencephalitis in 1 case and brain abscess in the other, have also been reported (35).
In a study in which one group of mice was infected with A fumigatus alone, another given corticosteroids in addition to being infected with Aspergillus, and other groups infected with Aspergillus and given either TNFα or anti-TNFα with or without corticosteroids, no deaths occurred among mice given TNFα, compared with 40–80% mortality in the other groups (36). TNFα-treated mice also had fewer organs infected with Aspergillus. One proposed mechanism to explain the role of TNFα in preventing invasive aspergillosis is through enhanced antifungal activity of polymorphonuclear phagocytes (37). The percentage of polymorphonuclear cell (PMN)–induced hyphal damage was increased 30 minutes after incubation of the PMNs with TNFα. TNFα also increased the production of superoxide anion by PMNs in response to nonopsonized hyphae. Another study, in which both neutropenic and non-neutropenic mice were challenged with intratracheal A fumigatus, demonstrated an increase in lung TNFα levels (38). This was associated with the development of a patchy, peribronchial infiltration with mononuclear cells and PMNs. The administration of antibodies directed against TNFα led to an increase in mortality in both normal and neutropenic mice, in addition to an increase in the fungal burden. There was decreased neutrophil influx in both groups of mice treated with anti-TNFα. The administration of a TNFα agonist peptide to the neutropenic mice 3 days prior to infection with Aspergillus conidia resulted in an increase in the survival rate from 9% to 55%.
The occurrence of invasive aspergillosis in a patient receiving infliximab for Crohn's disease was reported. Of note, it occurred 5 days after administration of the first dose of infliximab, and the patient was not receiving other immunosuppressive therapy (14). Allogeneic bone marrow transplant patients with grade III–IV graft-versus-host disease who received infliximab had a significantly higher risk of invasive aspergillosis (6 of 11 patients) than did patients with similar graft-versus-host disease who did not receive infliximab (4 of 41 patients) (39). The authors of that report recommended that preemptive antifungal therapy directed against filamentous fungi should be strongly considered for this high-risk population.
Neutralization of TNFα in vivo resulted in increased H capsulatum colony-forming units and 100% mortality in naive and immune mice challenged with H capsulatum administered intranasally (40). H capsulatum may cause subclinical infection or flu-like illness with or without bronchopneumonia in immunocompetent subjects. Patients with marked immunosuppression often present with progressive disseminated histoplasmosis, which is associated with significant mortality. There have been reports of disseminated histoplasmosis following exposure to infliximab (41) and etanercept (42). In the MedWatch voluntary reporting system, 10 patients who developed histoplasmosis were reported, including 9 treated with infliximab and 1 with etanercept (1). In the patients who received infliximab, symptoms of histoplasmosis occurred within 1–24 weeks of the first dose. All of the patients lived in areas endemic for Histoplasma. Nine of 10 patients were treated in intensive care units, and 1 died. Since the publication of the initial FDA report, an additional 12 cases of histoplasmosis associated with anti-TNF therapy have been reported, including 10 in patients treated with infliximab and 2 in patients treated with etanercept (41). Four of these patients died (3 infliximab treated, 1 etanercept treated).
TNFα has been demonstrated to play a key role in a murine model of pulmonary C neoformans infection. Neutralization of TNFα has prevented pulmonary clearance in mice with pulmonary cryptococcosis (43), possibly through inhibition of cell-mediated immunity. There have been 2 reports of cryptococcal pneumonia in patients with underlying RA who received infliximab therapy, including 1 patient who developed cryptococcal pneumonia after 3 infliximab infusions (44, 45). Several important points regarding TNF blockade can be extrapolated from these case reports. The first is that common illnesses, such as community-acquired pneumonia, may be caused by atypical organisms. Second, in immunocompromised hosts receiving TNF blockade treatment who do not respond to initial empiric antimicrobial therapy, a definitive tissue diagnosis is required. Finally, invasive cryptococcal disease of the lungs can occur despite negative results of tests for serum cryptococcal antigen.
Pneumocystis jiroveci (formerly, Pneumocystis carinii).
TNFα is important in host defense against P jiroveci. Severe combined immunodeficient mice that acquire pulmonary infection with Pneumocystis are able to clear the organism within 19 days after reconstitution with splenic cells from immunocompetent mice. Treatment of reconstituted mice with anti-TNFα IgG almost completely inhibited clearance of Pneumocystis from the lungs (46). The finding of TNFα in lung homogenate supernatants from reconstituted SCID mice further supports the significance of endogenous TNFα in the clearance of pulmonary pneumocystosis. There have been at least 17 cases of Pcarinii pneumonia (PCP) in patients receiving TNFα blockade therapy; however, the majority of these occurred in patients who were receiving other immunosuppressive agents including methotrexate and/or corticosteroids, which could themselves predispose to Pneumocystis pneumonia (1).
While therapy with TNF blockade has led to dramatic clinical improvement in certain inflammatory diseases, there has been a price tag associated with its use—typical and atypical infections. Although pulmonary TB and extrapulmonary TB have received the most attention, a wide range of infectious diseases have been observed in the setting of TNF blockade treatment. Since the biologic effects of TNFα on immune cells are widespread, it is not surprising that TNF inhibition has resulted in a decreased ability to control infection, which has been demonstrated in both animal models and human studies. The therapeutic prescription that takes into account the net state of immunosuppression of the patient as well as a rational antimicrobial program is a key component of the management of diseases to be treated with anti-TNF inhibition. As we learn more about infectious diseases associated with TNF blockade, the therapeutic prescription will continue to evolve.